vs
Moonshot AI vs StepFun
Two Chinese open-weight labs at opposite scales. Kimi K3 is a 2.8T flagship for hard coding; Step 3.7 Flash is a small-active model for cheap vision work.
By The Subconscious Team · Updated
Moonshot AI vs StepFun: key differences
Moonshot and StepFun both release open weights and both build vision into their models, but they aim at different jobs. Kimi K3 is a 2.8 trillion parameter mixture-of-experts model with a 1M window, priced at $3 in and $15 out, and it posts near-frontier coding results. Step 3.7 Flash is a 198B mixture-of-experts vision-language model with only 11B active parameters and 256K context, priced at $0.20 in and $1.15 out. StepFun's models trail frontier systems on hard multimodal reasoning, so the price gap buys a real capability gap.
Self-hosting shows the difference most clearly. K3 needs a 64+ accelerator cluster, and its custom license adds a commercial agreement above $20M in hosting revenue. Step 3.7 Flash ships under Apache 2.0 and runs on vLLM and SGLang, and its small active parameter count keeps hosting cheap. StepFun also builds speech and audio models. K3's drawbacks are speed and verbosity. StepFun's are thin Western distribution and support, with China-hosted first-party inference, though OpenRouter carries both labs. For repo-scale coding, K3 wins. For vision and video understanding in cost-sensitive agents, Step 3.7 Flash does.
What Moonshot AI and StepFun do
Moonshot AI
Moonshot AI is the Beijing lab behind the Kimi models. Its flagship Kimi K3 launched July 16, 2026 as a 2.8 trillion parameter mixture-of-experts model that activates 16 of 896 experts per token, with native vision and a 1M token context. It is the first open model in the 3T class, and full weights landed on Hugging Face on July 27. The hosted API costs $3 in and $15 out per million tokens, with cached input at $0.30, and it runs through an OpenAI-compatible endpoint, Kimi Code in the terminal, OpenRouter and Cloudflare Workers AI.
Example models: Kimi K3, Kimi K2.6
Full Moonshot AI profileStepFun
StepFun is a Shanghai AI lab known for efficient multimodal models, with a mix of proprietary API models and open-weight releases. Its current workhorse, Step 3.7 Flash, came out in May 2026 as a 198B mixture-of-experts vision-language model with only 11B active parameters. It has 256K context, selectable reasoning levels, tool use and structured outputs, and it ships under Apache 2.0. StepFun's own API prices it at $0.20 in and $1.15 out per million tokens, and OpenRouter carries it too.
Example models: Step 3.7 Flash, Step3
Full StepFun profileShould you choose Moonshot AI or StepFun?
Moonshot AI
Choose Moonshot AI for
- Hard repo-scale coding where capability matters most
- Tasks that need 1M context rather than 256K
- Teams that want the strongest open model
StepFun
Choose StepFun for
- Image and video reading at $0.20 per million input
- Easy self-hosting with Apache 2.0 weights
- High-volume multimodal calls at $0.20 in
Moonshot AI vs StepFun at a glance
| Attribute | Moonshot AI | StepFun |
|---|---|---|
| Model access | Open weights, custom license | Open (Apache 2.0) and API models |
| Flagship models | Kimi K3, Kimi K2.6 | Step 3.7 Flash, Step3 |
| Speed | ~33 tok/s on Kimi K3 | ~128 tok/s on Step 3.7 Flash |
| Price | $3 in, $15 out (Kimi K3) | $0.20 in, $1.15 out (Step 3.7 Flash) |
| Customization | Open weights to fine-tune | Open weights to fine-tune |
| Deployment | API, Kimi Code, OpenRouter | First-party API, OpenRouter |
| Long context | 1M | 256K |
Frequently asked questions
What is the difference between Moonshot AI and StepFun?
Two Chinese open-weight labs at opposite scales. Kimi K3 is a 2.8T flagship for hard coding; Step 3.7 Flash is a small-active model for cheap vision work.
When should I choose Moonshot AI over StepFun?
Hard repo-scale coding where capability matters most; Tasks that need 1M context rather than 256K; Teams that want the strongest open model.
When should I choose StepFun over Moonshot AI?
Image and video reading at $0.20 per million input; Easy self-hosting with Apache 2.0 weights; High-volume multimodal calls at $0.20 in.
Is Moonshot AI or StepFun cheaper?
Moonshot AI: $3 in, $15 out (Kimi K3). StepFun: $0.20 in, $1.15 out (Step 3.7 Flash). The cheaper choice depends on the model and workload.
Which has more context, Moonshot AI or StepFun?
Moonshot AI: 1M. StepFun: 256K.
Related comparisons
Running long-horizon agents?
If your agents run past 200K tokens, compare both against Subconscious. Our inference stack treats a long-horizon trace as the primary workload, so speed, cost, and accuracy hold up deep into the trace.